CAREER: Future of Search: User, Social Networks and Language

如何在特定的环境下检索特定用户的相关信息是一个具有挑战性的研究问题。这项研究项目的目标就是应对这一挑战，为下一代搜索引擎奠定基础。这里开发的方法不是简单地将查询与包含查询词的文档匹配，而是一个统一的以用户为中心的检索框架，包括：(1)个性化：学习考虑用户的内容、上下文和决策标准的用户模型；(2)语言处理：学习更好的文本表示，以便从不同的语料库和语言资源中进行检索；以及(3)社交网络：基于社会规范和用户的社交网络进一步改进用户模型。为了评估该框架，将开发一个个性化的社交搜索引擎。该项目的结果将是一个统一的检索框架，采用一套适用于各种信息检索任务的新技术，包括：搜索引擎、推荐系统和自适应过滤系统。通过PI的行业合作，该项目的成果有望被纳入商业系统，从而使数百万用户受益。硅谷的K-12学生、本科生、研究生和工程师将参与并受益于该项目。该项目将产生以研究为基础的教育材料(讲座幻灯片、视频讲座、书籍章节和课程项目)。项目成果，包括出版物、开放源码软件、注释数据集、演示和课程材料，将通过项目网站(http://www.soe.ucsc.edu/~yiz/futureofsearch).传播

How to retrieve relevant information for a specific user under a specific set of circumstances is a challenging research problem. The goal of this research project is to tackle this challenge and lay the foundation for the next generation of search engines. Instead of simply matching a query to documents that contain query words, the approach developed here is a unified user-centric retrieval framework that consists of: (1) personalization: learning a user model that takes into consideration the content, context and decision criteria of a user; (2) language processing: learning better text representation for retrieval from heterogeneous corpus and linguistic resources; and (3) social networks: further improving the user model based on social norms and a user's social networks. To evaluate the framework, a personalized social search engine will be developed. The result of this project will be a unified retrieval framework with a set of novel techniques applicable across a wide range of information retrieval (IR) tasks, including: search engines, recommender systems and adaptive filtering systems. Through the PI's industry collaboration, the results of this project are expected to be incorporated in commercial systems, and thus benefit millions of users. K-12 students, undergraduate, graduate students, and engineers in Silicon Valley will be involved and benefit from the project. The project will lead to research-based educational materials (lecture slides, video lectures, book chapters and course projects). Project results, including publications, open source software, annotated data sets, demos and course materials, will be disseminated via the project website (http://www.soe.ucsc.edu/~yiz/futureofsearch).